Why an Offline-First Learning System Can't Be Agnostic About Content

 Imagine a hospital built for a disaster zone. Its engineers have solved extraordinarily hard problems: a power grid that holds when the regional grid fails, a water system that stays clean when supply lines are cut, redundant generators, backup batteries, and careful load management so the whole facility does not collapse the first time demand spikes. It is, by every infrastructural measure, an excellent hospital. Now imagine that same hospital's leadership says, when asked what medicine is stocked on its shelves, "That's not really our department. We built the systems that keep medicine cold and available. What medicine ends up in the building is a pharmaceutical question, not a hospital infrastructure question."

This is an absurd position for a hospital to take (although, maybe not from a U.S. healthcare system perspective), and it is absurd for a specific reason: the power grid and the water system were never the point. They were built in service of the medicine. A hospital with flawless infrastructure and an empty or arbitrary pharmacy has not achieved neutrality. It has failed at the very thing the infrastructure was meant to support.

This is the position Offline-First Design for Learning (OFDL) occupies with respect to educational content.

OFDL, developed by Marc Alan Sperber through work in refugee camps and other low-resource, crisis-affected contexts, solves real problems with real rigor. It calls for servers that run on under twenty watts, sustained by solar panels and battery packs, so that learning continues through power outages. It insists that all course materials, including videos, handouts, and articles, be integrated directly into the learning management system rather than dependent on external links, because a link to the outside world is a link to a connection that may not exist. It specifies open standards so that content is not hostage to a single vendor's offline capabilities. It builds in two-way synchronization so a learner's progress survives the transition between offline and online, and between one location and the next.

This is the equivalent of the hospital's power grid: hard won, well designed, and absolutely necessary. The problem is not that this infrastructure exists. The problem is what it leaves unexamined.

OFDL says nothing about where its content comes from. It does not distinguish between Open Educational Resources, which are freely available and openly licensed for reuse and adaptation, and proprietary content, which is controlled by publishers, corporations, or certifying bodies that can restrict, revoke, or price access at will. It treats content as a payload to be compressed, resized, cached, and synchronized.

Educational content, however, is not merely something to be delivered. It often must be transformed. In many of the low-resource and crisis-affected communities OFDL is designed to serve, learners speak languages that are largely invisible to commercial educational publishers. A lesson available only in English or French may technically exist on the server, yet remain educationally inaccessible to the children sitting in front of it.

For many kinds of infrastructure, agnosticism would be reasonable. A shipping container does not need an opinion about what is inside it. A road does not care whether the truck is hauling grain or luxury goods. OFDL is different because it does not claim to be neutral about outcomes. Its own stated principles, including a holistic learning experience that guarantees equal opportunity for engagement, localized digital literacy built for a learner's specific environment, and continuity across the disruptions that make low-resource and crisis-affected communities vulnerable, are explicit commitments to educational equity. Whether those commitments are realized depends overwhelmingly on the content learners receive.

A brilliantly engineered offline server delivering content licensed to a single grant cycle or content that disappears the moment funding ends has not achieved educational continuity. Neither has a server stocked with proprietary materials that cannot legally be translated, revised, or culturally adapted for the communities it serves. By contrast, a server distributing Open Educational Resources (OER) provides content that can persist independently of any publisher's business model while empowering local educators to translate it, revise it, remix it, and improve it without seeking permission from distant rights holders.

In multilingual classrooms, refugee settlements, Indigenous communities, and rural schools, translation is not a convenience. It is the difference between participation and exclusion. A lesson that cannot be translated is, for many learners, no lesson at all. A science lesson translated into a child's first language does more than improve comprehension. It communicates that the learner's language and culture belong in the classroom. The ability to localize educational materials is not an implementation detail. It is one of the principal mechanisms through which educational equity is achieved. Translation transforms access into understanding, and understanding into learning. Openly licensed educational resources make that transformation legally and practically possible.

A framework that meticulously specifies its technical architecture yet remains silent about whether educators possess the legal freedom to translate and adapt the very materials they deliver has mistaken the visibility of its engineering for the completeness of its mission.

Infrastructure and educational content are not separate concerns that happen to occupy the same system. The infrastructure exists because the content matters. An offline-first learning framework can only claim to advance educational equity honestly if it has something meaningful to say about both.

OFDL has solved the engineering problem of keeping learning available during disruption. But availability alone is not educational access. Without a commitment to openly licensed, adaptable, and translatable content, the framework cannot fully realize its own goal of equitable learning. Infrastructure makes learning possible. Open, adaptable, and translatable content makes learning accessible. A framework that aspires to educational continuity cannot remain silent about content.

Sticky Fingers and Literacy Tax Dollars?

This is the time of year when Minnesota state legislators have town halls and fund raisers. At one recent event, I asked a long term state legislator if they knew how much of the money allocated by the READ Act that they had just bragged about went into the coffers of private equity companies. The response included a mumble about unintended consequences. 

State literacy initiatives are presented as investments in children. Increasingly, they have also become intentional, consequential investments for private equity (PE.) In both Mississippi and Minnesota, a small group of investment firms now owns many of the products that schools are required, or strongly encouraged to purchase. The same firms profit from screening students, training teachers, and renewing those contracts year after year. The same firms have their fingers in the legislative money pots of a lot of states.

Curriculum Associates (i-Ready), Renaissance Learning (FastBridge and Star), Edmentum (Exact Path), NWEA (MAP Growth), and Lexia's LETRS teacher-training program represent the most widely used literacy products in these states. Behind these familiar education brands are a handful of private equity firms including Veritas Capital, Blackstone, Francisco Partners, Permira, Berkshire Partners, Hellman & Friedman, and Vistria Group.  These money makers have consolidated ownership across the literacy marketplace.

This matters because these are not one-time purchases. Student screeners require annual licenses. Teacher certifications require ongoing training and renewal. Every school year, public education dollars flow back through the same recurring-revenue business model. A note here about the use of screens in school: identifying literacy needs increasingly routes through licensed software platforms, sometimes student-facing computer programs, sometimes tablet apps that teachers use to record one-on-one assessments, but in either case, the data and scoring run through proprietary, licensed vendor software rather than locally-built and locally owned and controlled programs. I guess cuz Minnesotans are smart enough to do that kind of thing..

A growing share of the infrastructure surrounding public literacy policy is controlled by financial firms whose legal obligation is to maximize returns for investors. When the same investment groups own the assessments that identify literacy needs and the professional development used to address them, policymakers and taxpayers should ask whether sufficient safeguards exist to ensure that educational decisions are driven by evidence rather than by market concentration.

The issue is not whether these products are effective, although that's still a very open question. The issue is accountability. When state policy effectively channels recurring public dollars into a marketplace dominated by a small number of financial firms, citizens deserve transparency about who owns these products, how much public money is spent on them each year, how those contracts are awarded and renewed, and whether meaningful competition still exists, and whether the programs are actually working and not being massaged as has been widely reported about Mississippi’s numbers.

 Generating dependable returns for investment portfolios should not be the object of legislators. 

The Fear-Driven QA Trap

 When institutions or governments respond to AI-generated OER content with centralized quality gatekeeping, they're operating from a scarcity and control mindset, the same mindset that OER is designed to move away from. The logic goes: "AI tools are producing low-quality materials at scale, so we need stronger filters before anything reaches learners." That feels responsible, but it smuggles in some problematic assumptions.

The radical promise of OER isn't just free resources; it's freedom to adapt. The 5Rs framework (Retain, Reuse, Revise, Remix, Redistribute) places the teacher at the center as an active agent, not a passive consumer of vetted content. A teacher who finds a resource that is 80% right for their classroom isn't supposed to wait for a quality-assured version; they're supposed to fix the 20%.This means imperfection is not the enemy of OER; it's assumed. The whole architecture of OER anticipates that materials will need local adaptation. When quality assurance becomes the primary response to AI-generated OER content, several things happen that cut against OER's grain:

 It re-centralizes authority. A review board or quality committee becomes the arbiter of what's usable, recreating the gatekeeping dynamic of traditional publishing. That’s comforting to bureaucrats but not useful for teaching and learning.

It signals distrust of teachers. The implicit message is that educators can't judge whether a resource suits their students, that they need experts to pre-approve it for them.

It slows the ecology down. OER's strength is its velocity and adaptability. Heavy QA processes introduce bottlenecks that favor static, "finished" resources over living, iterable ones.

It mistakes polish for fitness. A highly polished resource that doesn't fit a specific classroom context is less useful than a rough one that a teacher can quickly reshape. QA processes typically optimize for the former.

Rather than asking "is this resource good enough?" before release, the OER-consistent question is "does this resource come with enough transparency for a teacher to assess and adapt it?" That shifts the work from pre-publication gatekeeping to:

Providing Clear metadata about how, when, and with what tools a resource was made;

Creating in Editable formats like Moodle courses so adaptation and localized assessments are actually possible, not just theoretically permitted. These formats need to be able to be used equally in environments with the least resources as well as classrooms with the latest technology.

Providing a Platform for Community Annotation so teachers can flag issues and share improvements contextually;

Build Educator Capacity that helps teachers develop the critical eye to evaluate AI-generated content themselves.

Using fear of low-quality AI content as a reason to reduce teacher agency is contradictory to what gives OER its beauty. Teachers know their students and are always going to be a better quality filter for their specific context than any generalized review process. Fear-based QA trades that distributed, context-sensitive intelligence for a centralized standard that fits every classroom approximately and none just right.

The better bet is trusting the OER ecosystem's own immune system which includes: teachers adapting; communities annotating; and bad resources simply not getting used or getting improved rather than building walls at the gate. Spend money on teachers in classrooms not filter creators.